Build multitask recommenders on GCP

Data Structure:

• user model(query model): using features: userId and organization
• article model(candidate model): contentId
• mutlitask recommender:
  • framework: tensorflow-recommender
  • predicting:
    1. predict time spend on pages
    2. predict article wachtes

File Strcture:

• serving: CI/CD; kubflow; ai-platform; bigquery; tensorflow-recommenders

  

• application: serving model predicting API with google app engine

  

• streaming: dataflow; bigquery;

  

Global environment:

Change below environment variables into your version:

• model name: tfrs
• model version: mf
• project name: buddie-270710
• region: west-europe4

Source:

Modified from here